phoenix-oss/llama-stack-mirror
0
0
Fork 1
mirror of https://github.com/meta-llama/llama-stack.git synced 2025-12-28 02:41:59 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

Added proper support for calling tools without parameters.

Browse source
This commit is contained in:
ilya-kolchinsky 2025-05-09 12:38:59 +02:00
parent 7a8e398f09
commit 3be85c717f
2 changed files with 85 additions and 31 deletions
Download patch file Download diff file Expand all files Collapse all files

67
llama_stack/providers/remote/inference/vllm/vllm.py
View file

@ -166,47 +166,52 @@ def _process_vllm_chat_completion_end_of_stream(
) -> list[OpenAIChatCompletionChunk]:
chunks = []
args_str = tool_call_buf.arguments
args = None
try:
args = {} if not args_str else json.loads(args_str)
except Exception as e:
log.warning(f"Failed to parse tool call buffer arguments: {args_str} \nError: {e}")
if finish_reason is not None:
actual_finish_reason = _convert_to_vllm_finish_reason(finish_reason)
else:
actual_finish_reason = StopReason.end_of_message
if args:
chunks.append(
ChatCompletionResponseStreamChunk(
event=ChatCompletionResponseEvent(
event_type=current_event_type,
delta=ToolCallDelta(
tool_call=ToolCall(
call_id=tool_call_buf.call_id,
tool_name=tool_call_buf.tool_name,
arguments=args,
arguments_json=args_str,
if tool_call_buf.tool_name:
# at least one tool call request is received
args_str = tool_call_buf.arguments or "{}"
args = {}
args_parsed_successfully = True
try:
args = json.loads(args_str)
except Exception as e:
args_parsed_successfully = False
log.warning(f"Failed to parse tool call buffer arguments: {args_str} \nError: {e}")
if args_parsed_successfully:
chunks.append(
ChatCompletionResponseStreamChunk(
event=ChatCompletionResponseEvent(
event_type=current_event_type,
delta=ToolCallDelta(
tool_call=ToolCall(
call_id=tool_call_buf.call_id,
tool_name=tool_call_buf.tool_name,
arguments=args,
arguments_json=args_str,
),
parse_status=ToolCallParseStatus.succeeded,
),
parse_status=ToolCallParseStatus.succeeded,
),
)
)
)
)
elif args_str:
chunks.append(
ChatCompletionResponseStreamChunk(
event=ChatCompletionResponseEvent(
event_type=ChatCompletionResponseEventType.progress,
delta=ToolCallDelta(
tool_call=str(tool_call_buf),
parse_status=ToolCallParseStatus.failed,
),
else:
chunks.append(
ChatCompletionResponseStreamChunk(
event=ChatCompletionResponseEvent(
event_type=ChatCompletionResponseEventType.progress,
delta=ToolCallDelta(
tool_call=str(tool_call_buf),
parse_status=ToolCallParseStatus.failed,
),
)
)
)
)
chunks.append(
ChatCompletionResponseStreamChunk(